Advanced Technology for On-Demand Logistics
The sharing economy has taken the world by storm. As more consumers and companies opt for services on demand, logistics-on-demand providers like Lalamove are reaping the last mile opportunities. Based in Hong Kong, Lalamove’s app connects users across 22 international markets, with over 7 million registered users and 700,000 drivers among its ranks.
Expansion has been the company’s primary strategy since its launch in 2013. But with a firm foothold in many countries, and to keep up with changes brought on by events in 2020, Lalamove is shifting its focus away from operations to advanced technology applications like machine learning (ML) and serverless computing. The business has been using Amazon Web Services (AWS) since 2016 as its primary cloud provider.
Striving for Oversupply of Drivers
Customer satisfaction on the Lalamove app largely hinges on the availability of drivers and vehicles to match logistics needs. Samy Basset, machine learning engineer at Lalamove, explains, “From a data science perspective, the more drivers we have, the more efficient our matching engine will be in matching drivers and customers.” Lalamove strives for at least 10 available drivers in close proximity per order to support its app’s ability to meet each customer’s specific requirements.
According to Basset, competition for drivers in the markets in which Lalamove operates can be fierce. On average, there are two to four companies in the market vying for on-demand logistics orders and drivers to fulfill them. A fast onboarding process is key to recruiting more drivers. “Our app and analytics optimization activities won’t work if we don’t have enough supply. And the smoother onboarding is, the more likely we can increase our pool of drivers,” he says.
Automating Review of Drivers’ Licenses
Lalamove began a project using optical character recognition (OCR) in 2020 to automate, and accelerate, the process for checking drivers’ licenses. At the first stage of onboarding, drivers must upload pictures of their driver’s license to the Lalamove app. Previously, Lalamove employees would review the photos and determine whether to accept the document and progress to the next phase of onboarding, or decline it due to poor image quality.
The manual process was time-consuming, often requiring up to a week to verify or reject a driver’s ID. In case of rejection, the driver would receive a message and have to resubmit their ID photo, further delaying their start date for work. This could cause applicants to become frustrated and exit the application process, seeking work elsewhere.
Saving Human Resources and Cost
Lalamove evaluated three open-source and public-cloud OCR tools. The engineering team ultimately chose Amazon Textract because it was easy to use, required no ML experience, and performed well at recognizing poor-quality documents. They were also attracted by the key value detection features, which recognize certain fields including date of birth and driver’s name.
The support received from the AWS team in implementing OCR for the first time also swayed Lalamove’s decision. The data and engineering team was able to implement a solution within a couple of weeks, whereas Basset estimates it would’ve taken 3–4 months to build an end-to-end OCR solution in house. “Automating this step in the onboarding pipeline with Amazon Textract is saving us both human resources and cost. Amazon Textract has enabled us to scale and expand into new markets faster, as we no longer have to hire additional staff to perform manual document review,” Basset adds. Lalamove has reduced document processing costs by around 10 times compared to manual OCR workloads.
Accept/Reject Message Sent within 5 Seconds
Lalamove rolled out the automated OCR in Manila first, with plans to extend utilization to its operations in Jakarta and Kuala Lumpur soon after. The solution developed on Amazon Textract is easily replicable and scalable, for use in any market with English-language drivers’ licenses. Plus, the solution is serverless and integrates well with other technologies in use at Lalamove, such as AWS Lambda for serverless orchestration of code and Amazon API Gateway for serverless API calls.
Furthermore, this vital first onboarding step is completed in near real time. Within 5 seconds, drivers receive an automated response indicating acceptance or rejection of their ID. Total driver onboarding time has fallen from weeks to 1 hour as a result. “We’ve improved the user experience, as drivers now know if they can start working with us right away,” Basset says.
Managed Services Simplify ML Applications
In addition to Amazon Textract, Lalamove is adopting other AWS ML services to boost automation. It’s now using Amazon Rekognition to detect how many vehicles with a promotional sticker from Lalamove or one of its competitors pass through a busy intersection at a given time. This analysis helps the company determine its market share.
Lalamove has also used
Amazon SageMaker to build a data science workbench, with a toolset, ML model library, and established data pipeline to facilitate ML experimentation and development. Data scientists are currently trying to scale up new automation features, for which the compute power and ease of building and striking scenarios in Amazon SageMaker are well suited.
“It’s convenient to get the ML resources we need on-demand, and it makes experimentation cost effective,” Basset says. “We have a limited capacity in terms of data scientist resource hours. With ML tools on AWS, we can prototype ideas quickly and reduce our time-to-market by relying on managed services instead of developing in-house solutions.”
Proactive Support for Expansion
Lalamove plans to continue rolling out new features and app enhancements in the markets in which it operates and is now building the business in Asia and South America. ML, including OCR technology, will play an increasingly important role in Lalamove’s operations.
The company appreciates the proactive support it continues to receive from AWS in terms of suggesting technologies to facilitate expansion and providing solution architects and data scientists to support prototyping and implementation. “If we use AWS solutions to automate more workflows like driver onboarding, this will improve our capacity for innovation and certainly benefit our end customers. Automation is a game changer,” Basset concludes.
About Lalamove
Lalamove provides on-demand, same-day, and advanced-order delivery services by connecting more than 7 million registered users with delivery drivers on its mobile and web apps. The company operates in 22 markets across the globe.
Benefits of AWS
- Responds to driver uploads in 5 seconds
- Reduces costs by tenfold compared to manual OCR
- Saves 3–4 months’ development time
- Facilitates onboarding in new markets
- Simplifies ML development and testing
- Lowers time-to-market for ML initiatives
AWS Services Used
Automatically extract printed text, handwriting, and data from any document
Learn more »
Amazon Rekognition
Automate your image and video analysis with machine learning
Learn more »
Amazon Sagemaker
Build, train, and deploy machine learning (ML) models for any use case with fully managed infrastructure, tools, and workflows
Learn more »
Get Started
Companies of all sizes across all industries are transforming their businesses every day using AWS. Contact our experts and start your own AWS Cloud journey today.
